The overarching goal of this renewal application is to develop and disseminate novel methods of design and analysis for effectiveness research using methodology for adaptive treatment strategies (ATSs). An ATS is a dynamic algorithm for matching clinical treatment decisions to the evolving course of the individual patient's response to treatment over time, based on a list of rules that together specify the sequential, multi-stage decision making of the clinician treating a chronic disorder such as depression. The classic randomized controlled trial has been augmented to include patient and clinician preferences (Equipoise Stratification, or ES), and extended to the Sequential Multiple Assignment Randomization (SMAR) trial, a design for comparing ATSs. Nonetheless, the 'fixed- treatment' trial continues to be a mainstay of much of psychiatric research. The 'fixed' study protocol precludes adaptive changes to treatment for patients who do not fully respond to assigned treatment, typically leading to problematic rates of nonadherence and dropout. The uncontrolled treatment decisions that are intrinsic to nonadherence jeopardize the power and interpretability of intent-to-treat (ITT) comparisons, but reflect a lack of fit between clinical reality and the fixed-treatment design, rather than a weakness of ITT. The proposed aims extends ES-SMAR to (1) develop designs to close the gap between ITT inference and clinically relevant inference, using ATS that preempt nonadherence, (2) develop methods to address the external validity of ITT comparisons based on multi-stage trials of ATS, using single-stage effectiveness trials and well designed observational studies for calibration. Both mathematical analysis and simulation will be used to develop and evaluate the new methods, building on recent progress in the area. The methods will also be tested on data from a large-scale effectiveness trial (Sequenced Treatment Alternatives to Relieve Depression - STAR*D) sponsored by the National Institute of Mental Health. The proposed new methods will be developed in the specific clinical context of mental disorders, but they are clearly relevant to many other areas of medicine. PUBLIC HEALTH RELEVANCE: Nonadherence to prescribed medication and therapy pervades the treatment of most psychiatric disoders. It complicates experimental evaluation of new treatments, but because nonadhernce may lead to worse patient outcomes, it also complicates decision making by the clinician in practice settings. New approaches to clinical trial design that preempt or reduce nonadherence in the experiment and the real world are needed. This proposal will help develop the new design for experiments and the methods needed for their analysis